keras implement of blockQNN -- Practical Block-wise Neural Network Architecture Generation exclude the Distributed Asynchronous Framework
original paper 'Practical Block-wise Neural Network Architecture Generation' by zhaozhong et al. is available at here
ps. Since the Distributed Asynchronous Framework is not implemented, I had test the code and run for several epoch just for a temporal model, but if you want to complete the whole training progress, it's may take a couple of days, which depends on the GPU cards.